Behavioral science Study Guide

📖 Core Concepts Behavioural Science – The interdisciplinary study of human behaviour, combining psychology, neuroscience, genetics, economics, sociology, etc. Behaviourism – A psychological movement focused on observable actions and their environmental determinants. Operant Conditioning – Learning through consequences (reinforcement / punishment). Classical Conditioning – Learning by association between a neutral stimulus and an unconditioned stimulus. Cognitive Biases – Systematic shortcuts (e.g., loss aversion, anchoring) that shape decisions; exploited in nudges. Reinforcement‑Learning Model – Formal framework where agents update value estimates based on prediction errors. Bayesian Decision Framework – Agents combine prior beliefs with new evidence to compute posterior probabilities and choose actions. Neuroimaging Modalities – fMRI (spatial), EEG (temporal), MEG (spatial + temporal). --- 📌 Must Remember Key Figures & Contributions Skinner – Operant conditioning, schedules of reinforcement. Pavlov – Classical conditioning (CS‑US pairing). Watson – Early experimental behaviourism. Bandura – Social learning (observational learning, self‑efficacy). Core Definitions – Behavioural science = study of human behaviour; integrates natural & social sciences. Typical Experimental Tools – Operant boxes, lesion studies, intracranial electrodes, fMRI, EEG, MEG. Behaviour Change Levers – Defaults, framing, loss aversion, reminders, social norms. RL Update Rule – $V{t+1}=Vt+\alpha\,(Rt-Vt)$ (learning rate $\alpha$, reward $Rt$). Bayes’ Rule – $P(H|D)=\dfrac{P(D|H)P(H)}{P(D)}$ (posterior ∝ likelihood × prior). --- 🔄 Key Processes Classical Conditioning Present neutral stimulus (NS). Pair NS with unconditioned stimulus (US). After repeated pairings, NS → conditioned stimulus (CS). CS alone elicits conditioned response (CR). Operant Conditioning (Skinner Box) Subject performs a response (e.g., lever press). Immediate consequence delivered (reinforcer or punisher). Reinforcement schedules (fixed‑ratio, variable‑interval) shape response rate. Reinforcement‑Learning Cycle Agent observes state $st$. Chooses action $at$ based on policy $\pi(s)$. Receives reward $rt$ and next state $s{t+1}$. Updates value: $Q(st,at) \leftarrow Q(st,at)+\alpha\,[rt+\gamma\max{a'}Q(s{t+1},a')-Q(st,at)]$. Designing a Nudge Identify target behaviour. Diagnose relevant bias (e.g., present‑bias). Choose lever (default, framing). Test with A/B experiment; iterate. --- 🔍 Key Comparisons Operant vs. Classical Conditioning Operant: behaviour produces consequence → learning about effects. Classical: behaviour elicited by stimulus → learning about associations. fMRI vs. EEG fMRI: high spatial resolution (mm), low temporal (seconds), measures BOLD. EEG: high temporal resolution (ms), low spatial resolution, records electrical activity. Reinforcement Learning vs. Bayesian Decision RL: learns from reward prediction errors; model‑free or model‑based. Bayesian: updates beliefs using probability calculus; optimal under known priors. --- ⚠️ Common Misunderstandings “Behaviourism ignores mental processes.” – Modern behavioural science often integrates cognition (e.g., cognitive‑behavioural models). “All neuroimaging shows causation.” – fMRI/EEG are correlational; causal inference requires lesions, TMS, or stimulation. “A bias is always irrational.” – Biases are adaptive heuristics; they become problematic only in certain contexts. “Reinforcement always strengthens a behavior.” – Reinforcement can be positive (adding) or negative (removing) but both increase likelihood; punishment typically decreases it. --- 🧠 Mental Models / Intuition “Behaviour = Stimulus + Response + Consequence” – Treat any observable action as a loop: what triggers it, what it is, and what follows. “Bias as a shortcut shortcut.” – Imagine a mental “fast lane” that shortcuts detailed calculation; nudges reroute traffic onto the fast lane. “Brain → Behavior ≈ Software → Output.” – Neuroimaging tells us which “modules” are active, not the exact “code”. --- 🚩 Exceptions & Edge Cases Schedule of Reinforcement – Variable‑ratio schedules produce the highest, most resistant‑to‑extinction response rates (e.g., gambling). Neuroimaging – fMRI BOLD signal can be confounded by vascular changes; not a direct neural firing measure. Bayesian Updating – If prior is extremely strong, new data may have minimal impact (priors dominate). --- 📍 When to Use Which Choose Classical vs. Operant – Use classical when studying stimulus–response associations (e.g., phobias). Use operant for voluntary actions shaped by outcomes (e.g., habit formation). fMRI vs. EEG – Use fMRI for locating where activity occurs (e.g., brain region of reward). Use EEG for when events happen (e.g., timing of error detection). Reinforcement‑Learning Model – Ideal for tasks with clear reward feedback and trial‑by‑trial learning. Bayesian Model – Preferred when prior knowledge is strong and uncertainty quantification matters (e.g., diagnostic decision‑making). --- 👀 Patterns to Recognize “Reward → Increased response rate” – Look for reinforcement contingencies in experimental descriptions. “Loss framing → Stronger health‑behaviour change” – Health‑intervention questions often hinge on loss aversion. “Default option = Majority choice” – In consumer or policy scenarios, the default is a powerful predictor. “Variable‑interval schedules → steady but low response” – Contrast with variable‑ratio (high, bursty). --- 🗂️ Exam Traps Confusing “positive reinforcement” with “positive punishment.” – Both add something; only reinforcement increases behavior. Assuming fMRI shows causation. – Remember it’s correlational; the correct answer will emphasize “association”. Mixing up “bias” vs. “heuristic.” – Bias = systematic error; heuristic = mental shortcut that can be unbiased. Selecting “fixed‑ratio” when the question describes “high, resistant‑to‑extinction” behaviour. – That describes variable‑ratio. Choosing “Bayesian” for a task that only gives immediate reward feedback. – Reinforcement‑learning is the appropriate framework. ---